The No.1 Website for Pro Audio
Does repeated destructive editing of lossless file degrade the file?
View Poll Results: Will this degrade the sound?
No, why would it?
7 Votes - 53.85%
Yes, of course.
6 Votes - 46.15%
Voters: 13. You may not vote on this poll

Old 19th August 2018
  #1
Does repeated destructive editing of lossless file degrade the file?

(To be clear, I'm not asking if repeated copying of a WAV degrades the file. I know the answer to that is no.)

So, if I paste a bunch of WAV masters into, let's say, SoundForge, to sequence an album after mixing and mastering....And I do dozens of tiny cuts in volume, raises in volume, removing small spaces, adding silence between tracks, AND I HIT "SAVE" AFTER EACH little edit, will it degrade the file?

I know people will say "you should do that in a project file of a non-destructive NLE program", but I prefer destructive editing for this. Particularly Sound Forge.

If I do 120 cuts and saves in a 90-minute long, 24-bit, 44.1k stereo WAV file, will it degrade the overall audio at all?

I Googled a bunch of things and couldn't find the answer anywhere, at least not from people who seemed to know what they were talking about.

I figured y'all here would know for sure one way or the other.

Thank you!
=-=-
EDIT: I think this has been answered, Answer for SoundForge seems to be "No, you're all good, audio wise, doing this. Though you won't have many levels of un-do."

See my conclusions here:
Does repeated destructive editing of lossless file degrade the file?
based on PraxisAxis telling me how to do a Null Test:
Does repeated destructive editing of lossless file degrade the file?

Also see the comment by stinkyfingers explaining why there was one piece of odd data in my tests:
Does repeated destructive editing of lossless file degrade the file?

=-=-=-

Last edited by SyntheticThought; 20th August 2018 at 09:59 AM.. Reason: adding that I think this has been answered, because people are still responding without reading the thread. (as we all do.)
Old 19th August 2018
  #2
Lives for gear
 

Quote:
Originally Posted by SyntheticThought View Post
So, if I paste a bunch of WAV masters into, let's say, SoundForge, to sequence an album after mixing and mastering....And I do dozens of tiny cuts in volume, raises in volume, removing small spaces, adding silence between tracks, AND I HIT "SAVE" AFTER EACH little edit, will it degrade the file?

I know people will say "you should do that in a project file of a non-destructive NLE program", but I prefer destructive editing for this. Particularly Sound Forge.
I bolded the answer to your question.

If you save over the source file, there is no source file anymore.
Old 19th August 2018
  #3
Lives for gear
 

Quote:
Originally Posted by SyntheticThought View Post
If I do 120 cuts and saves in a 90-minute long, 24-bit, 44.1k stereo WAV file, will it degrade the overall audio at all?
And that is another question.
Most modern DAWs operate internally at 64bit, so if you save to a file with high enough bit depth, then the answer is: no.

Then if you do not only cuts and pastes, but also some volume edits and save to a low bit depth file (we could argue if 24bit is low), the answer is: yes.
Old 19th August 2018
  #4
Lives for gear
 
Meriphew's Avatar
 

Everytime you edit a digital file some of the 1’s and 0’s fall off into the recycle bin.
Old 19th August 2018
  #5
Lives for gear
 
daviddever's Avatar
Quote:
Originally Posted by SyntheticThought View Post
(To be clear, I'm not asking if repeated copying of a WAV degrades the file. I know the answer to that is no.)

So, if I paste a bunch of WAV masters into, let's say, SoundForge, to sequence an album after mixing and mastering....And I do dozens of tiny cuts in volume, raises in volume, removing small spaces, adding silence between tracks, AND I HIT "SAVE" AFTER EACH little edit, will it degrade the file?

I know people will say "you should do that in a project file of a non-destructive NLE program", but I prefer destructive editing for this. Particularly Sound Forge.

If I do 120 cuts and saves in a 90-minute long, 24-bit, 44.1k stereo WAV file, will it degrade the overall audio at all?

I Googled a bunch of things and couldn't find the answer anywhere, at least not from people who seemed to know what they were talking about.

I figured y'all here would know for sure one way or the other.

Thank you!
Some context would be appropriate here.

In general, you shouldn't be re-sequencing a record after the mastering process, and especially, while using a destructive editor.

Track-to-track leveling should be done during the mastering process (as that's really the point); any changes should be spooled out ("rendered") after the finished product is approved by the client. NEVER, ever destructively edit your source files unless they're in hopelessly bad shape, and then only copies thereof.

Think of the mastering process as a continuous operation performed across a set of tracks, with the finished artifact being a tagged, continuous audio image (loosely speaking, referencing Red Book CD) plus any additional metadata required for your release format.
Old 19th August 2018
  #6
Lives for gear
 
JayTee4303's Avatar
I don't know the full answer, but I know this.

A DAW project is a collection of "files".

There are audio data files, and there are processing instructions, and these two are seperate.

Destructive editing saves an altered audio file. Degrade... improve... beauty is in the eye of the beholder, but altered IS altered.

Non-destructive editing saves the original audio file "as is", along with altered processing instructions.

Rendering, bouncing, etc, saves an altered audio file.

In theory, a word processing document can serve to illustrate.

Open it, change a word, save it. That word is changed. The others are not. A destructive edit, but no degradation of the un-edited words. Identical ASCII 8 bit groups of ones and zeros.

Save it one billion times, and those ASCII codes won't change, ever. Except the ones YOU changed.

The theory, and my knowlege break down if the DAW rounds things off, or lossy compresses audio files at each save.

I don't know about that.

I save EVERY project, every session, every major edit, as an incrementally numbered Cakewalk Bundle.

Each numbered file has its OWN copy of every audio file. Wave files.

I don't hear any issues, and generally end up with 20-40 saved bundles, per finished project. More with demanding clients.
Old 19th August 2018
  #7
Lives for gear
 

The problem with your question is that “degrade” has no meaning here. When people talk about degrading audio files, they usually mean “making changes that shouldn’t alter the sound (like converting between formats) but actually do alter the sound, in a way I consider to be bad”.

You’re explicitly talking about changes to the file that alter the sound. Only you can decide whether the change has “degraded” the file.
Old 19th August 2018
  #8
Lives for gear
 

Yes, too much cutting is bad for the cutting tool. It gets less sharp and so the edges you cut are going to get a bit fuzzy. Depends on the material you cut if thats negligible or an issue. Percussive material is more prone to deteriorate than for example warm lush pads.
Old 20th August 2018
  #9
Quote:
Originally Posted by whiteaxxxe View Post
Yes, too much cutting is bad for the cutting tool. It gets less sharp and so the edges you cut are going to get a bit fuzzy. Depends on the material you cut if thats negligible or an issue. Percussive material is more prone to deteriorate than for example warm lush pads.
First part of this sounds like a joke answer. Last sentence sounds plausible.
Old 20th August 2018
  #10
Thank you for the answers here.

Most of these sound reasonable, even when entirely contradicting answers by other people here.

This is the quandary I was in when I posed the question.
Old 20th August 2018
  #11
TJT
Lives for gear
Quote:
Originally Posted by daviddever View Post
Some context would be appropriate here.

In general, you shouldn't be re-sequencing a record after the mastering process, and especially, while using a destructive editor.

Track-to-track leveling should be done during the mastering process (as that's really the point); any changes should be spooled out ("rendered") after the finished product is approved by the client. NEVER, ever destructively edit your source files unless they're in hopelessly bad shape, and then only copies thereof.

Think of the mastering process as a continuous operation performed across a set of tracks, with the finished artifact being a tagged, continuous audio image (lossely speaking, referencing Red Book CD) plus any additional metadata required for your release format.
Agreed. And even if you haven't had it mastered yet, it's a really bad idea to treat full tracks like this. Even if you're most comfortable working in a single waveform editor, you might make editing decisions that you will regret the next day, or next week. Even if you have the original source file, you won't have your edit history, so you'll have to start all over again and you might forget some of the decisions made. Best to re-orient yourself and get comfortable doing it in a non-destructive way.
Old 20th August 2018
  #12
Lives for gear
 
Praxisaxis's Avatar
 

Quote:
Originally Posted by SyntheticThought View Post
Thank you for the answers here.

Most of these sound reasonable, even when entirely contradicting answers by other people here.

This is the quandary I was in when I posed the question.
You can scientifically prove the answer (whatever it is) to yourself, without relying on anyone else's advice, by using a null test. This will give you a 100% definite indication, one way or the other.

Look up Null Test if you don't know how to do it (it's easy). Then get your original audio, put a heap of cuts in, but no actual changes so that it remains time-locked with the original (if that makes sense). Then do the null test: ie invert the audio, put it in another track to play back with the original, and you should get exactly no signal (no sound), if and only if your edits did not degrade (ie., change) the original audio.
Old 20th August 2018
  #13
Quote:
Originally Posted by Praxisaxis View Post
You can scientifically prove the answer (whatever it is) to yourself, without relying on anyone else's advice, by using a null test. This will give you a 100% definite indication, one way or the other.

Look up Null Test if you don't know how to do it (it's easy). Then get your original audio, put a heap of cuts in, but no actual changes so that it remains time-locked with the original (if that makes sense). Then do the null test: ie invert the audio, put it in another track to play back with the original, and you should get exactly no signal (no sound), if and only if your edits did not degrade (ie., change) the original audio.

Praxisaxis:

PERFECT.

Thank you. Will do soon, and will report back.

Before I do, would this be a good way to do it?:

Make a file with a null set, then an hour of music, then another null set? And only edit the volumes of various parts of the music, up and down?

Then only look for silence replacing the null set parts after I phase out at the end?

Because I'm assuming here that
1. The volume changes in the music won't phase out to null, i.e that there WILL be something there after the phase mix at the end.

2. I"m also assuming that if there IS artifacting, editing the music should also affect the file in the null set sections, even if I don't edit those parts. Right?
Old 20th August 2018
  #14
Quote:
Originally Posted by Praxisaxis View Post
......Then do the null test.....

I actually thought of a bonehead variation on this, but Null Test is much better.

If I couldn't find a definitive answer, or a way of proving it definitively myself (which I believe the Null Test should do), I was going to do this;

Make a file with a bunch of sine waves (but only one frequency, not a bunch of frequencies), do edits on some, then compare screenshots of both in layers in Photoshop, reducing the opacity. Then try that at different zoom levels on the time line, and see if I can spot a difference.

But Null Test sounds worlds better. And no need to reinvent the go kart when there's a perfectly good plane I can use free.
Old 20th August 2018
  #15
Lives for gear
 
Praxisaxis's Avatar
 

To be honest, I'm not really sure what you're testing for, because I haven't got my head around whatever it is you're asking.

My vague understanding is that your questions can be condensed to this:
Will save + save + save (etc) = change to the audio data? As I understand it, the answer is "no" but you can test this easily using a null test.
Will edit + save + edit + save (etc) = change the audio data? Again, I think the answer is "no," apart from the very thing that you are changing with your edits. If I was going to test this using a null test, I would place a bunch of false edits, which constitute an edit, but do not apparently change anything. So, for example, place a cut, reduce that section by - 5 decibels, hit save, increase the same section by 5 decibels, hit save. Run this cycle as many times as you want, for veracity, or do similar things in different parts of the audio (e.g. reverse a section, hit save, then reverse it again, etc, etc). Then run your null test, putting your experimental audio alongside the original audio.
Old 20th August 2018
  #16
Quote:
Originally Posted by Praxisaxis View Post
To be honest, I'm not really sure what you're testing for, because I haven't got my head around whatever it is you're asking.

My vague understanding is that your questions can be condensed to this:
Will save + save + save (etc) = change to the audio data? As I understand it, the answer is "no" but you can test this easily using a null test.
Will edit + save + edit + save (etc) = change to the audio data? Again, I think the answer is "no," apart from the very thing that you are changing with your edits. If I was going to test this using a null test, I would place a bunch of false edits, which constitute an edit, but do not apparently change anything. So, for example, place a cut, reduce that section by - 5 decibels, hit save, increase the same section by 5 decibels, hit save, then run your null test.


Cool.

Will it provide data as useful if the sine set is a continuous sweep? There are presets to generate in Sound Forge.

Or does it need to be discreet tones with silence between them?
Old 20th August 2018
  #17
Lives for gear
 
Praxisaxis's Avatar
 

Quote:
Originally Posted by SyntheticThought View Post
Cool.

Will it provide data as useful if the sine set is a continuous sweep? There are presets to generate in Sound Forge.

Or does it need to be discreet tones with silence between them?
Not sure what you're asking. I'd just try it on any old audio. When you create the "null," what you're looking for is *any* signal at all (it may not be very loud, so you might have to use some kind of metering).

Try doing a null test before you do anything else, to confirm the method. Because you're working with digital audio, you should find that when you invert your sound, then play it at exactly the same time as the original, there is literally no remaining signal - silence plays back.

If you want to get a sense of how it works, try applying a small amount of reverb to some audio, then doing the null test with the original. You'll find you are left with only the small amount of reverberation - it is entirely accurate.
Old 20th August 2018
  #18
Lives for gear
 
Praxisaxis's Avatar
 

p.s. have another look at my earlier post, I edited it to be a bit clearer.
Old 20th August 2018
  #19
Quote:
Originally Posted by Praxisaxis View Post
Not sure what you're asking. I'd just try it on any old audio. When you create the "null," what you're looking for is *any* signal at all (it may not be very loud, so you might have to use some kind of metering).

Try doing a null test before you do anything else, to confirm the method. Because you're working with digital audio, you should find that when you invert your sound, then play it at exactly the same time as the original, there is literally no remaining signal - silence plays back.

If you want to get a sense of how it works, try applying a small amount of reverb to some audio, then doing the null test with the original. You'll find you are left with only the small amount of reverberation - it is entirely accurate.


Got it. I'm on it. Control group tests worked perfectly. Resulted in silence.
Thank you!
Old 20th August 2018
  #20
Lives for gear
 
Praxisaxis's Avatar
 

Quote:
Originally Posted by SyntheticThought View Post
Got it. I'm on it. Control group tests worked perfectly. Resulted in silence.
Thank you!
No probs. curious to no know what you come up with. My guess is (if you follow the method I outlined a couple of posts up), you're going to get silence.
Old 20th August 2018
  #21
Quote:
Originally Posted by Praxisaxis View Post
p.s. have another look at my earlier post, I edited it to be a bit clearer.
Quote:
Originally Posted by Praxisaxis View Post
p.s. have another look at my earlier post, I edited it to be a bit clearer.


SCIENCE, BEOOTCHIES!

This is brilliant.

Here's my control group. All I changed was raised and lowered volumes on the two parts you see here. Will move on to actual tests now.

You so smart. Woot!
Can't yet embed images here, but this is a screenshot:
https://biptunia.com/wp-content/uplo...08/Capture.jpg
Old 20th August 2018
  #22
Quote:
Originally Posted by Praxisaxis View Post
To be honest, I'm not really sure what you're testing for, because I haven't got my head around whatever it is you're asking.

Basically I'm asking "Is there any reason why I shouldn't sequence a record with destructive editing, if I keep backups, because that's how I like doing things. And my sequencing involves tiny bits of raising and lowering songs, or parts of songs, in the master single audio file, as I go."

(I release songs as complete long album-length single files, aka "Dark Side of the Moon", but not as good. lol.)

That's how I compose records, and present them, to be listened to from start to finish. I later also release as individual songs, but I take those songs out of my master album-length file, with transitions intact, so it sort of works on Spotify or an iPod, or any player that plays a playlist of individual songs in order. Because not everyone wants to download a 1 gb FLAC torrent, or a 200 meg 320k MP3 in one shot. And a lot of phones have trouble navigating files like that, at least if you hit "pause".)
--
SO, I did some tests, and wanted to run the interesting / confusing results by you before I make 100 more edits and kill my carpel tunnels tonight.

Screenshot:
https://biptunia.com/wp-content/uplo...intersting.jpg

This is null tests after phase swapping, then doing a bunch of destructive volume edits on one file, than summing them.

The left side of the image was a song, the right side of the image was a different song, the 3 sections in the middle were sine sweeps from too low to hear up through too high to hear.

The very middle section, the sine sweep between markers 1 and 3, was a control group where I did zero edits. It was, as I expected (and hoped), null after the test.

Everywhere I did a raise or lower of volume and then a save, you see what you'd expect. Non-null results. Good.

Everywhere I did nothing, you see what you'd expect, null, which is also good.

The only unexpected deviation is the thin lines of non-null on the far right: what I did there was lowered by 50%, using the Sound Forge built-in volume default
"Lower by 50% (6 db)".
Then hit "Save."

Then I RAISED the exact part by 200%, without moving the in and out points,
using the Sound Forge built-in volume default
"Raise 200% (6 db)".
Then hit "Save."

I also tried the same thing, but with picking in and out points as close to zero crossing on both channels as I could. (zero crossing on one channel, very close to same on second. I couldn't find 2 places with zero crossing on both channels for in and out.)

The results were the same.

--Is this a problem indicating there IS destruction of the audio while making these edits?

--Or is it just that maybe the Sound Forge presets are not perfect? (I couldn't manually enter the amounts, SF won't let you type in values, and the slider won't get you exactly to 200 or to 50. Very close, but not dead on.)

--Or is this me just not really understanding logarithms, and maybe I shouldn't have daydreamed in 9th grade math class and said to myself "How is this won't help me in later life? I'm gonna ROCK!"

Thank you for your service!
MWD
Old 20th August 2018
  #23
Lives for gear
 
stinkyfingers's Avatar
 

I would suspect sound forge is not using the correct values for 50% and 200%.
Those should be “bit perfect” gain changes.
Also, I would recommend using a good meter for testing (if not already) because most are low resolution and give false results. (Says -inf but is really not).
Old 20th August 2018
  #24
Gear Addict
Well, with utmost respect to the more qualified answers here and PraxisAxis in particular; for myself i´ve always been satisfied with using the following rules when saving:

* keep bit rate and sample rate
* do not ever save or export to a "lesser" or destructive format (i.e. wav to mp3)

Besides that i just let my ears decide usually. Works for me, but i am a mere hobbyist.
Old 20th August 2018
  #25
Quote:
Originally Posted by defdaft View Post
Well, with utmost respect to the more qualified answers here and PraxisAxis in particular; for myself i´ve always been satisfied with using the following rules when saving:

* keep bit rate and sample rate
* do not ever save or export to a "lesser" or destructive format (i.e. wav to mp3)

Besides that i just let my ears decide usually. Works for me, but i am a mere hobbyist.

Yup.

Problem is I have tinnitus from decades of rock bands in the past, plus a hole with a tube in it in one ear. It's a procedure called a myringotomy.

It's the most metal thing in the world, but doesn't make for totally reliable ears.

Trigger warning to anyone reading this: Do not do a Google image search for "myringotomy" if you are squeamish.
Old 20th August 2018
  #26
Lives for gear
 
Acid Mitch's Avatar
 

When you load a file all you are doing is making a copy in memory. When you save you are copying the data from memory to a file on disk. Unless you do something to degrade the contents, the data remains the same.
With word documents you can cut, paste, delete, edit spelling and sentences,etc to your hearts content with no degradation. It’s the same with audio files. You can move sections around, delete sections, add sections,etc with no degradation when saving.

If there was any degradation happening you would hear it and wouldn’t have to ask.
Destuctive editing just means that any changes to the file are permanent. It doesn’t mean sound quality or anything gets changed.
Old 20th August 2018
  #27
Lives for gear
 
Praxisaxis's Avatar
 

Quote:
Originally Posted by Acid Mitch View Post
When you load a file all you are doing is making a copy in memory. When you save you are copying the data from memory to a file on disk. Unless you do something to degrade the contents, the data remains the same.
With word documents you can cut, paste, delete, edit spelling and sentences,etc to your hearts content with no degradation. It’s the same with audio files. You can move sections around, delete sections, add sections,etc with no degradation when saving.

If there was any degradation happening you would hear it and wouldn’t have to ask.
Destuctive editing just means that any changes to the file are permanent. It doesn’t mean sound quality or anything gets changed.
AFAIK that is correct and what I'd have expected. But it doesn't hurt to do things the long way around and actually perform a test. Strictly speaking, though not what we'd expect, it would not be impossible for some software (I don't know what the OP is using) to do something unexpected on editing/saving. Science sometimes yields unremarkable results .
Old 20th August 2018
  #28
Lives for gear
 
Praxisaxis's Avatar
 

Quote:
Originally Posted by SyntheticThought View Post
Trigger warning to anyone reading this: Do not do a Google image search for "myringotomy" if you are squeamish.
*Immediately googles. . .*
Old 20th August 2018
  #29
Quote:
Originally Posted by Praxisaxis View Post
*Immediately googles. . .*

Lol. Reminds me a bit of when I interviewed Brendon Small, and mentioned the effects of Krokodil to him. He stopped mid-interview on a live terrestrial radio show to Google it. He was horrified.

I was honored to have been able to shock the creator of Dethklok.

But yeah, I have a hole & a tube in my left eardrum. Not nifty. lol.

Is your user name a nod to Laswell? Or a libertarian thing?

MWD
Old 20th August 2018
  #30
Quote:
Originally Posted by JayTee4303 View Post
IBM PC/XT w 8086 main and 8087 math co-processor
1 MB RAM
5 MB Seagate Hard Drive
SoundBlaster Pro Interface
Cakewalk 2.0 by Twelvetone


Is this your legit gear? Or a joke.

If not a joke, I'm intrigued at the use of vintage, antique computer gear. What's the reason?

📝 Reply
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
🖨️ Show Printable Version
✉️ Email this Page
🔍 Search thread
🎙️ View mentioned gear
Forum Jump
Forum Jump